| Title: | VAXELN |
| Notice: | Support - 2.*; Kits - 3.* |
| Moderator: | HELIX::MIANO LO |
| Created: | Fri Feb 08 1991 |
| Last Modified: | Wed Jun 04 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 770 |
| Total number of notes: | 2721 |
I'm a little bit confused as I'm requested to try solving a customer
VAXELN problem whereas I did not work with ELN for a looong time
(several years)... SO if somebody could help...
The customer problem is the following:
- he runs VAXELN-V4.3 (old version !) on a RTXAV300 system
- he states that when he send messages on Ethernet AND on the VME bus
(at the same time) the system hangs after a while. He states that before
the system hangs he can watch at the system behavior by Edisplay and he
does not see any resource problem or else. He says that when the system
runs normaly he has about 12 EZDRIVER processes but the system hangs as
a process 15 or 16 appears (always EZDRIVER). When that occurs then
Edisplay is hung and the customer is disconnected.
How can we know which is this process 15 or 16 ? Is there a way to
debug this problem ? Debugging the EZDRIVER ?
- if messages are sent EITHER on Ethernet OR on VMS Bus the problem
never appears.
Any help would be greatly welcome as I'm lost ....
Thanks in advance.
Jean-Pierre
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 767.1 | Some thoughts to try | ZYDECO::BODA | Realtime Expertise Center | Tue Apr 22 1997 16:20 | 38 |
Hi Jean-Pierre,
Yep, V4.3 is a vintage version!
Not sure I know an answer, but here are a few questions I would ask:
1) What is the rtVAX-300 in? A KAV30, or third party like the AEON
VME300?
2) Have they written their own VMEbus backplane network driver? Or
are they simply writing data to the VMEbus at the time they are doing
a ker$send to a remote system across the ether?
3) Do they have the local debugger built in? I would make sure they
do as this may give you a better glimpse at exceptions raised while
at elevated ipl. Make sure there is a physical console attached.
4) The KAV30 support was integrated into VAXELN V4.4 and had numerous
changes made to the KAV30 kernel at that time. (Before then there
was the optional KAV30 Toolkit). I know changes were made to AST
handling and the like.
5) Does anything else continue to run at the time of the "hang"?
6) Is it visible on the network (e.g., NCP> TELL mynode SHOW EXEC)?
7) Are they using Internet services? If so, can they ping the node?
8) Are they elevating IPL anywhere in their code?
9) Does a VMEbus analyzer show anything abnormal when they do
writes to the VMEbus?
Perhaps some of these ideas may shed some more light on the problem.
Regards,
Alan
| |||||
| 767.2 | PRSSOS::VOUTERS | Philippe VOUTERS, Evry (France) | Wed May 07 1997 06:55 | 36 | |
Solution :
Edit ELN$:EZPROCESS.PAS
const
.
.
.
datalink_desc_max = (rcv_desc_max * 10) + xmt_desc_max;
--
.
.
.
Search procedure transmit_message
Replace :
if desc = nil
then
goto return;
by
if desc = nil
then
begin
allocate_buffer(desc, cb);
if desc = nil
then
goto return;
end;
And this does the trick for the customer (the target has run over 22
hours without failures). They stress the EZDRIVER, hence the 10
multiplier in datalink_desc_max.
Best regards,
Philippe
| |||||